Based over these sources

https://github.com/vcg-uvic/viper [github.com]

I've implemented first attempt

https://www.patreon.com/posts/56970585 [www.patreon.com]

If someone have intention to test it - you're welcome PM me - sorry, windows binaries only available now.
It will not commercial product later, I will publish sources after polishing and all licensing terms will be solved.

Keep in touch, it evolves daily now.

Thank you.

https://www.patreon.com/posts/57010321 [www.patreon.com]

Rod bunches dynamics added.

https://www.patreon.com/posts/57058430 [www.patreon.com]

Implemented special datatype for contraction controlling per-muscle. It lacks originally at Viper system sources.

https://www.patreon.com/posts/57059475 [www.patreon.com]
And about collision support (remember, that no inbetween "pills" are visible still. A bit later)

https://www.patreon.com/posts/57427411 [www.patreon.com]
progress at this moment : consistent frame tracking implemented - now it is possible to use strand as an object - and SDF collision processing is added too (for nodes only now, for pills - possible later) - it is new for VIPER system

https://www.patreon.com/posts/57499195 [www.patreon.com]
Next iteration - more sophisticated contraction control, have ideas about reimplementing it as CVEX procedure

https://www.patreon.com/posts/57581379 [www.patreon.com]
1. Figured with cross-section shape-matching constraints, Viper provides two kind of them, one is faster, but weak with bending - another slower and much more robust and plausible. Now all muscle data have per-muscle controls about shaping using and its type. Since this constraint is really slow - and not so redundant as bend or stretch - so it should be processed in different way, than more computationally cheap. It lacks parallelization of computation - and there is some possibilites for it. Will research it a bit later. I have plans to cret such membrane-contraints with more control - with ability to add it everywhere it may be needed by rigger and configured more freely.
2. Contraction and hardening computation moved to CVEX procedures and it gets more comfortable to setup such phenomena - and have more Houdini's style. To be true, hardening is not implemented still, I have no ideas about it. Yes it is about straightening of bend-constraints rest pose, but it is buried deep into bend constraint implementation and overall solver concept.

Next step - make capturing and mesh movement.

https://www.patreon.com/posts/57747223 [www.patreon.com]
Finished deformer subsystem. It uses harmonic capturing (whatever it can mean), and it works plausible.
Some issues introducing shape-matcher constraint, so I decide to substitute it with triangle matching, which works much faster and better parallelized. And it required more sophisticated base rods geometry preparation, I seem, but I have sufficient ideas about implementation.